Evaluating a Trading Strategy
We have spent the last few lessons defining what a trading strategy is. You might have noticed that a strategy can be built out of millions of parameters, which means there are billions of possible strategies out there. Mathematically it looks impossible to decide which ones are any good.
Believe me, it is not that difficult.
But we do need a scientific way to decide, rather than a feeling.
Let's start with common sense. For us to call a strategy good, it should make us money. We would not call a strategy good if it loses money.
So the decision looks simple. The strategy that makes more money is probably the better one.
But how do we know whether a strategy will make money? The only truly correct way is to define it, start trading it with real money, and find out. What else is there?
That method works. It is also expensive, and it is probably not the right thing to do.
So we use a proxy instead. It is called backtesting, and it is one of the most used words in trading.
What a backtest actually is
A backtest is a simulation of a trading strategy on past data. It answers one question: how much money would I have made if I had started trading this strategy some time ago?
That is all it is. A hypothetical simulation, run over data that has already happened.
It does not guarantee future returns. Nothing does. But once the simulation has run you get results, and you can read those results and make an informed decision about what to expect from the strategy from here.
A backtest does not promise you the same outcome tomorrow. It lets you make today's decision on something better than a hunch. The numbers that come out of it are statistically meaningful as long as the sample is large enough, and that is what lets you say whether the strategy has an edge.
It matters even when someone hands you the strategy
Assume someone you trust tells you about a strategy that makes money. Without really knowing the strategy, you just buy the stock.
Now what?
How much should you expect to make? Where do you get out if it goes against you? When does this strategy usually fail?
You have no answers, because nobody ever looked. A backtest is what gives you the target and the stop to place alongside the trade, and it tells you how strong your chances actually are.
Past performance is odds, not a promise
It is like expecting a good sportsperson to do well again today.
In football, if you had to put money on someone scoring, you would take Ronaldo or Messi over a player who has never scored in his life. In basketball you would put it on LeBron.
You are not certain. You are playing better odds. That is exactly the claim a backtest makes, and it is the only claim it is entitled to make.
Test every instrument on its own
Common sense also tells us to run the backtest separately for every instrument.
Every stock behaves differently and has its own technical character. A strategy that works well on one stock does not mean it will work on another. The same rules applied to a slow, heavy stock and to a fast, jumpy one are two different strategies in practice, and they deserve two different sets of numbers.
In the next lesson we start reading what a backtest hands back: the numbers on the scorecard, and which of them actually deserve your attention.